AutoGPT

Data poisoning is the latest threat for generative AI models

As AI gains traction, data poisoning emerges as a critical threat. Forcepoint identifies three attack types: availability attacks impairing entire models, backdoor attacks tainting training data, and targeted attacks on specific data subsets. The research also distinguishes between black-box attacks by outsiders and white-box attacks by insiders, highlighting the growing challenges in AI security. Source: […]